(107) Bridging the Gap and Ensuring Quality: Enhancing Real-World Data through Quality Control Implementation of an African Patient Population

Background: African populations are underrepresented in genome-wide association studies, making up less than 3%. The African Genomics Program (AGP), an initiative that aims to address the underrepresentation of African patients in Real-world Data (RWD) sets, aligning with Findable, Accessible, Interoperable, and Reusable (FAIR) principles. To bridge this gap, AGP is a collaboration with Discover Me South Africa DMSA, in this partnership longitudinal electronic health records (EHR) and genotype data from consenting patients attending Inkosi Albert Luthuli Central Hospital in Durban, South Africa are gathered. The data delivery is in tranches until the 100K patients target is achieved.

Objectives: To conduct thorough quality control assessment to ensure the suitability of the data, thereby minimizing research gaps in RWD and accelerating genomics research in Africa. To facilitate data integration the data was mapped to the OMOP common data model (CDM).

Methods: We defined objectives to ensure that the data addresses the gap in the patient population and meets our data standards and quality. These objectives served as the foundation for a comprehensive quality control assessment implemented collaboratively between the AGP and DMSA. For example, ensuring 90% of the population for each data tranche are of African descent and achieving 90% level of completeness for medication and hospital visits records. As part of the FAIR principles we mapped the data to the OMOP common data model.

Results: In total, 10 high-level data quality objectives were identified: 3 evaluating data completeness and patient ethnicity, 3 evaluating conformance to follow up time, 2 evaluating plausibility of the records and 2 evaluating genetic data quality and breast cancer, COPD and CKD therapeutic areas. Reproducible data quality dashboard and OMOP mapping scripts were developed. To map the data we took the tools and methodologies available to the OHDSI community. This process helped us identify terminologies in the dataset that could not be mapped to the standardised vocabularies.

Currently 25,392 EHR and ~5000 genotype patients’ data have been received in 3 data tranches. The hospital visit records span from 2002-2023. For the delivered data, 66.7% of the patients are of African descent, which is lower than the expected 90%. The diagnosis data is reported in ICD-10 codes and the medication data in a pseudo-RxNorm format. The OMOP mapping helped us identify codes that could not be mapped into the standardised format, for example, South African ICD-10 coding.

Conclusions: The process of generating and assessing the quality of data from underrepresented populations holds significant value. With this study we can ultimately enhance the real world evidence for advancing equitable and inclusive research.